[core] platform agnostic executor via collective_rpc (#11256)
Signed-off-by: youkaichao <youkaichao@gmail.com>
This commit is contained in:
@@ -862,12 +862,14 @@ def init_model_parallel_group(
|
||||
) -> GroupCoordinator:
|
||||
if use_custom_allreduce is None:
|
||||
use_custom_allreduce = _ENABLE_CUSTOM_ALL_REDUCE
|
||||
from vllm.platforms import current_platform
|
||||
return GroupCoordinator(
|
||||
group_ranks=group_ranks,
|
||||
local_rank=local_rank,
|
||||
torch_distributed_backend=backend,
|
||||
use_pynccl=True,
|
||||
use_custom_allreduce=use_custom_allreduce,
|
||||
use_pynccl=current_platform.is_cuda_alike(),
|
||||
use_custom_allreduce=current_platform.is_cuda_alike()
|
||||
and use_custom_allreduce,
|
||||
use_tpu_communicator=True,
|
||||
use_hpu_communicator=True,
|
||||
use_xpu_communicator=True,
|
||||
|
||||
Reference in New Issue
Block a user